I created a Claude Code skill to edit my videos

Setup

Hey everyone, this is Allan and today I’ll briefly show you how I edited yesterday’s video. Yesterday I tried something new: I asked Cloud Code to see if it could connect to Adobe Premiere and automate a few steps.

In my videos folder I keep all the recordings from each date. For the video from May 19 I have the screen capture, the webcam capture, and a separate audio track. I record with OBS, which saves the webcam video in one track and the screen video with the microphone audio in another. I keep the webcam and the audio separate because it’s easier to move the webcam around the screen when I edit.

Manual Editing

Normally in Adobe Premiere I load the two tracks, delete the video track that also contains the audio, and place the webcam track (which has no audio) on top. That gives me the screen audio from the first track and the webcam video from the second. That’s basically how I edit my videos.

AI Automation

I opened a terminal, navigated to the folder, and initiated Cloud Code to organize things. I used a shortcut that opens Cloud Code with the necessary permissions and told it to use the “video edit” skill on the May 19 recording. The skill launches a notebook that runs a Python script using the FFmpeg library.

The script first locates the video files and the webcam files based on the paths I gave it, then asks me what to name the final video. I chose “demonstration,” and the script starts running. It takes a little time, about five to ten minutes for a five‑minute video, which is roughly the same as my manual edit.

Script Details

I’m a front‑end developer, so I don’t know how to write FFmpeg code myself. The entire script was generated by AI. It imports the FFmpeg package, normalizes the audio loudness to the default level for YouTube or podcasts, sets the video quality, and then combines the screen video, the webcam video, and the audio into the final file. I have no idea how long it would have taken me to write this from scratch.

Outcome

While the script runs, I can read, study, or work on other things. When it finishes, the final video is ready. It took about seven minutes in total. I opened the original screen file and the webcam file, note that the webcam track has no audio. The final video contains both the screen and the webcam, with the audio boosted so my voice is clear.

This automation solves my video‑editing problem for now. I can record a video, ask AI to edit it, and get the finished product without opening Premiere myself. It makes creating YouTube videos much easier, and I’m happy with the workflow.

That’s it, guys. See you tomorrow.